Operator Variational Inference
Authors: Rajesh Ranganath, Dustin Tran, Jaan Altosaar, David Blei
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We study opvi with the Langevin-Stein objective on a mixture model and a generative model of images. 4 Empirical Study We evaluate operator variational inference on a mixture of Gaussians, comparing different choices in the objective. We then study logistic factor analysis for images. 4.1 Mixture of Gaussians 4.2 Logistic Factor Analysis ... Table 1: Benchmarks on logistic factor analysis for binarized MNIST. |
| Researcher Affiliation | Academia | Rajesh Ranganath Princeton University Jaan Altosaar Princeton University Dustin Tran Columbia University David M. Blei Columbia University |
| Pseudocode | Yes | Algorithm 1: Operator variational inference |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | We apply logistic factor analysis to analyze the binarized MNIST data set [24] |
| Dataset Splits | No | The paper does not explicitly provide details on validation splits or methodology. It focuses on training and testing on a held-out portion. |
| Hardware Specification | No | The paper does not specify the hardware used for the experiments. |
| Software Dependencies | No | The paper mentions 'Adam optimizer [12]' but does not provide specific version numbers for software dependencies. |
| Experiment Setup | Yes | We set the latent dimensionality to 10. ... The variational program generates samples by transforming a K-dimensional standard normal input with a two-layer neural network, using rectified linear activation functions and a hidden size of twice the latent dimensionality. ... For f , we use a three-layer neural network with the same hidden size as the variational program and hyperbolic tangent activations where unit activations were bounded to have norm two. ... We used the Adam optimizer [12] with learning rates 2 10 4 for f and 2 10 5 for the variational approximation. |